Back to Agent Zero

Browser Forms

plugins/_browser/skills/browser-forms/SKILL.md

1.131003 B
Original Source

Browser Forms

Use this skill for complex Browser form workflows where the page state may depend on selects, checkboxes, radios, file uploads, contenteditable fields, validation, or visual confirmation.

Start with browser:content to capture current refs, then use browser:detail on ambiguous fields before acting. Prefer ref-based form actions before coordinates.

Use select_option, set_checked, upload_file, type, type_submit, and submit for form interaction. Use coordinates only when no stable ref exists or the UI is intentionally canvas-like.

Use browser:screenshot plus vision_load when layout, visual validation, captcha-like UI, canvas content, or hidden state matters. Browser screenshots are not automatically loaded into model-visible history.

Verify after submission with browser:content, browser:state, or another explicit browser:screenshot plus vision_load.

Do not guess file paths for upload. Verify that every path exists before calling upload_file.